Flat-Earth: Quality Engineering (Step 4b)

Verdict: MINOR. No central verdict is likely to change, but several experimental records in the folder were never examined (they rest on digests only, or on no reading at all), one author is misidentified in the primer, and the arbiter and primer leave out the inventor's own rebuttal of the deciding 1870 test. These can be fixed by six deep reads and updates to facts.json, claims.md, arbiter.md (a "QE revision" section) and primer.md.

Tags: [read] = checked by me in the extracted text (_text/), a digest, or a saved file (page given); [calc] = my arithmetic (project venv, numbers shown); [memory] = recalled, confidence given. Quotations are short.

0. What this QE read

• Inputs: manifest.json (103 documents), facts.json (103 entries), claims.md, sites.md and the ten research notes, physicist.md, mathematician.md, contrarian.md, arbiter.md and primer.md in full; deep/ by section headings and where a check needed it.
• Digests (2.5 MB in total). Read in full: all 40 chunks of the 1893-97 Earth Review; Rowbotham 1865 (5 chunks); Carpenter 1885; Hampden 1869; Blount and Smith 1904; the Reddit and Steemit lists; every TFES page that was not deep-read (Aviation, Eratosthenes, FAQ, Maps, Moon, Sun, Voliva, Setting of the Sun, The Conspiracy); Heath, Russell, Earth.com, Kottke, FlatEarth.ws (two pages), AAP and Wikipedia; Skiba's non-deep-read pages and videos up to the Bauer debate. Read by targeted search (grep for experiments, measurements and the "flat"/"rotating" phrases, then the matching lines in context): Rowbotham 1881 (one 726,000-character chunk); the Airy 1856, Koresh 1898, IS-GPS-200N, NGA, DMA and Hirt digests; the remaining Skiba transcripts; and all 22 NASA, Army Research Laboratory (ARL) and Precise Time and Time Interval (PTTI) digests. I say which method I used wherever a finding depends on it.
• I ran example.py. It exits 0 and reproduces every number in primer section 5 (11.67 ft; k = 0.612 and 0.972; 6.00 ft sagitta; 78.2"/156.3"; H = 3,938/3,109/2,393 mi) [read: run output].

1. Coverage

Every manifest document has a digest (103 of 103; 40 of 40 chunks for the Review, 13 of 13 for IS-GPS-200N). No document was skipped in Step 0. The README (a finding aid) is correctly marked "private": true with a neutral author. No coverage defect in the manifest.

There is a gap outside the manifest. The Investigator saved documents to research/files/ that no later role read or summarised: - T. Westlake's letters to The Field (2 and 9 April 1870); - Carpenter's Field letters of 2 and 9 April 1870 (Field2, Field3); - Chambers's Journal (1870); - The Zetetic (July 1872); - Earth no. 51; - the Flat or Spherical? introduction; - the Flaming Sword issues of January to May 1897.

The WKU note lists them, but none is weighed in a review. Several bear on the 1870 wager (claim 1). This is a minor defect; see remediation item R7.

2. Shortcuts

3. Missed material (from the digests)

Ordered by weight for the claims.

Claim 1 (water level, no curvature) 1. Naylor's 1894 Bedford repeat (Review pp.216-217) [read: digest ch.33]. In August 1894 a zetetic party used a surveyor's theodolite, a Dumpy level and telescopes. They give measured bridge heights: Old Bedford Bridge 17 ft, Welney Bridge 13 ft, and the Great Eastern Railway bridge 19 ft to the top and 15 ft to the underside (from a company letter of 31 August 1894). They report Welney Bridge "a little below the cross-hair" once both levels were set, and the railway bridge visible as a dark band. They argue that on a globe the line of sight would pass 21 ft above that bridge. The variation they saw before levelling the transverse bubble is their own report of an instrument error. - The record matters for two reasons. It is the only post-1870 levelled-instrument test on the same stretch in the folder, and the physicist wrongly calls Naylor's work "cited, not held". - It is an interested-party account with no instrument readings in figures. It must be reported as the party's characterisation, weighed, and kept in the record. 2. The 1870 sketches reproduced (Review p.265): "true copies" of Carpenter's sketches from both ends, through an inverting telescope, with his Field report ("does NOT depart from a straight line"). claims.md says that "no neutral record of the signed sketches is in the folder". That is still true, since this is a partisan reproduction, but the folder does hold a copy, and the page image should be read (R1). 3. "Parallax Triumphant" (Review p.34) [read: digest ch.5]: - It quotes General Roy's trigonometrical-survey observation that thirty pickets 100 ft apart appeared in a right line in the afternoon but concave upward in the morning (Trig. Survey vol. 1, p.175). That is early documentary evidence of refraction varying through the day, which the arbiter's argument relies on without a historical source. - It also attributes to "Mr. Coulcher's sketches" a depression of "about 20 feet over six miles, not allowing for refraction". That needs checking against the Field text, because the arbiter states that no numerical offset was recorded. 4. Koreshan optical tests, 1896-97 (Koresh pp.84-106) [read: digest ch.03]. Globe requirements [calc]: - Summit canal, telescope 6 in, 5 mi: 11.4 ft hidden at k = 0; the 29-in target top needs k = 0.69 and the 7-in lower edge k = 0.87. - Naples Bay, telescope 2 in, 4.75 mi: the 2-ft target top needs k = 0.78 and its lower edge k = 0.93. - Gulf, telescope 1 ft, 6.5 mi: the 5-ft target top needs k = 0.63. These belong in the arbiter's "low-sight-line sightings" group, with the same verdict logic. The Naples Bay mid-point stake ("cross-bar ... in line with the target top", p.105) is described as a test of the middle of the line. Because the eye was 2 in above the water and the marks 2 ft, it is not a Wallace-type equal-height design; the arbiter should say so explicitly. 5. Campanella's four-mile light test (Behind the Curve, in Earth.com) [read: _text Earth.com]. Flat-earth proponents set "two styrofoam sheets ... at either end across a four-mile distance, with holes cut out at the same height". The light shone through one sheet "is not visible by Campanella". This is a near equal-height test run by the flat side, and it came out for curvature: the sagitta over 4 mi is 2.67 ft at k = 0 [calc]. The arbiter calls this item a "Knodel/Campanella ring-gyro" report, which merges two different experiments, and facts.json says the nature of the test is not described, but Earth.com does describe it. 6. Stretton 1905 (Ashby Canal) and Morrow 1896 (Wikipedia, which cites English Mechanic, 20 January 1905) [read: _text Wikipedia]. In Stretton's test a theodolite aligned with a boat's cabin roof showed "a dip ... of about eight inches" at one mile, against a tangent drop of 8.0 in [calc]. This is a secondary report, not in any review. The English Mechanic 1905 issues are a record to obtain. 7. The US Signal Corps heliograph of 1894, Mt Uncompahgre to Mt Ellen, 183 mi (Review pp.218-219; Harper's Weekly, 20 October 1894). The zetetic claim is that "nearly 6 miles" of curvature should hide the stations. - The tangent drop is 4.23 mi. - The horizon distances of peaks 14,418 ft and 11,410 ft high sum to 278 mi, more than the 183 mi between them. - The sight line passes about 7,330 ft above sea level at mid-path (k = 0) and 8,060 ft (k = 0.13), above the 7,000-ft plateau the article names [calc]. This is a clean "Refuted by calculation" example of the drop-from-sea-level error. It is missing from the reviews. 8. Middleton's plumb-line wire proposal (Review p.235, from 1871 letters): a non-optical design that would remove refraction from the question. It should be listed under "what would settle it" as the movement's own proposal.

Claim 2 (earth stationary)

9. Rowbotham's vertical-gun and air-gun tests (1865 pp.72-74; 1881 Exp. 2, book pp.66-67). He predicts 5,600 ft (1865) or 8,400 ft (1881) of lag. On a rotating earth the ball shares the gun's eastward speed, so the only lag is Coriolis: about 0.30 m west for a 16-s flight at 51.5° N [calc: Ω cos φ g T³/6]. That is far below wind and launch scatter for a ball fired upward. The experiment, if as described, therefore cannot detect rotation, and the 5,600-ft prediction is a frame error. His surface speed of 700 ft/s at London is also low: the true figure is 949 ft/s [calc]. The physicist analysed only Dubay's second-hand version of this anecdote.
10. Star-tube test of orbital motion (1865 p.75; 1881 Exp. 5). The claim that a star returns to "the same meridian time" after six months has to be read against sidereal time: a star transits about 3 min 56 s earlier each solar day [memory, high]. The deep read should establish what he actually claims.
11. Schoepfer's pendulum table (Review p.81; Carpenter p.40). The contrarian raises it ("could be a real ... effect", [memory, low-medium]) but neither the arbiter nor the primer carries it. It is minor.
12. Skiba's 2015 concessions (Part 2) [read: digest]: of a "Top 10" list he could not refute the Coriolis effect (#8) or the lunar eclipse (#2). The arbiter quotes his later "cannot prove any conspiracy" but not these concessions.

Claim 3 (small near sun and moon)

13. Rowbotham's sun triangulations (1865 p.79; 1881 book pp.102-104). For London and Brighton he gives noon altitudes of 61° and 64° over 50 mi, hence H ≈ 750 mi [calc: 751], but the latitude difference between the two places is only 0.685° [calc]. Either the altitudes were not observed simultaneously or they were misreported. The 1865 Ordnance Survey pair (London and Ackworth, 151 mi) differs by 2°16', which is close to the latitude difference, as a globe predicts [calc, to be confirmed by R3/R6]. This is the inventor's own data for the arbiter's "one H cannot fit" argument.
14. Naylor's concession (Review p.175, "Zetetic Refraction No. III") [read: digest ch.27]: sun angles along one meridian "correspond far more nearly" to the common theory than to Zeteticism. Together with G.M.'s table, this shows the movement's own writers conceding the sun-elevation test. It should be quoted in arbiter claim 3.
15. FE101's Indonesian Eratosthenes experiments (FlatEarth.ws, Skiba-Related; a critic's site): a flat-earth community "could not get a consistent figure for the Sun's distance". This is a secondary report of flat-side data and is not in any review.
16. Moonlight "cold light" claims: Rowbotham 1865 pp.104-107 (concentrated moonlight; the Lancet of 14 March 1856 reports "more than eight degrees" of cooling) and Skiba's own thermometer test (80.9 against 89.6 degrees, shadow warmer) [read: digests]. They are part of claim 3 (a self-luminous moon) and none of the three reviews gives them a verdict. A "Not testable as stated / Never properly tested" line is enough, with the radiative-cooling explanation marked [memory, medium].
17. Faulkner's star-trail analysis (Answers in Genesis, 2016, Skiba-Related) [read: digest]. It gives a quantitative comparison of a dome model with observed trails, with worked angles at 39° N. It is a critic's and an interested party's analysis, but pertinent to claim 3, and it is not in any review.

Claim 4 (gravity not a pull)

18. Airy 1856, Harton Colliery [read: _text p.35, p.37]. Over a depth of 1,256 ft "the acceleration of a seconds' pendulum below is 2s.24 per day", a gravity ratio below/above of 1.00005185 (p.43, digest). A uniformly accelerating floor predicts g h/c² = 4.2 × 10⁻¹⁴ [calc], so the observed change is about 1.2 × 10⁹ times larger. This is a [read] measurement that replaces the arbiter's and primer's [memory] gradient.
19. IS-GPS-200N [read: _text p.33; digest p.117, p.125]. The satellite clocks are offset by Δf/f = −4.4647 × 10⁻¹⁰ "for relativistic effects"; the specification gives Ω_e = 7.2921151467 × 10⁻⁵ rad/s and A_REF = 26,559,710 m. - Mass gravity (potential difference minus the velocity term) gives 4.457 × 10⁻¹⁰ [calc], a match to 0.2 percent. - A uniform field of strength g up to orbit gives 2.20 × 10⁻⁹ [calc], five times too large. - Kepler's third law with A_REF gives 11.966 h [calc]. This turns the physicist's [memory] "GPS period 11.97 h" into [read]. Note in the arbiter that the source is an interested party (the US GPS directorate).

People, history and minor accuracy items

20. Who was Oldham. He was H. Yule Oldham, a geographer at Cambridge [read: _text; facts.json; deep read]. The primer calls him "R. D. Oldham" twice, and the arbiter's records column sends the reader to the "Geological Survey of India context". That points to a different man (R. D. Oldham, the seismologist [memory, high]). See Corrections.
21. The academic record: Drayson's refusal and his list of tests in 1896 (Review p.197) and Proctor's 1873 concession (Myths and Marvels p.280, quoted on Review p.204 and p.218) that a sighting close to the canal surface would be decisive. Both belong in the arbiter's "dismissal without engagement" row (the Gilman, Baird and Proctor postscript row), which currently records only the 1885 non-replies.
22. The TFES Aviation page cites a 1979 New Scientist letter from the RAF College saying navigators are taught a fixed celestial sphere. It is pertinent to claim 5 as a quoted "admission" and is not addressed. The answer is the standard convention of the celestial sphere [memory, high].

4. Accuracy spot checks (against the extracted text)

5. Retention

1. Rowbotham's own rebuttal of the 1870 test (Experimental Proofs, 1870, research/files): the agreed three equal discs were not used ("an oblong flag" at the far end, nothing near the telescope), and refraction is among the instrument errors he names. - All three reviews carry it. The physicist calls it "a fair question of procedure", the mathematician a "fair design point". - Neither arbiter.md nor primer.md mentions it, although the strong-form verdict rests on that test. - The arbiter does keep Carpenter's "under protest" objection, so the omission is of the inventor's own voice. - Defect. It belongs in arbiter claim 1, "Where they differ", item 3, and in the primer's "Kept in the record" paragraph.
2. Koreshan Flaming Sword (March 1897) optics: a line of sight "curvilineating slightly upwards" through "gravic energy". The mathematician shows that this is the same 1/R upward-light rule as the TFES "Electromagnetic Acceleration". It is in the reviews but absent from the arbiter and primer. Minor: it is pertinent to the primer's "each model needs one light rule" lesson.
3. Goudey (1930) and Winship's Zetetic Cosmogony (1899, Durban) [Investigator leads; not read by anyone]. The arbiter's absence-of-evidence table does not list them as records held but not examined. Winship's book includes observations from Port Natal. Minor.
4. The libel record: Hampden's imprisonment in 1871 and his 1875 sentence. These are in the physicist's and contrarian's files. The primer keeps the civil ruling and "cheat and swindler are Hampden's words" but not the convictions. The record cuts against the flat side, so it should stay under SOURCE HONESTY. Minor.
5. Material the reviewers judged negatively but that was kept: the Clifton letter, the Knodel press report, the South Pole pendulum anecdote and the Koreshan bias question. No defect. The arbiter and primer keep all of them, stated fairly.

6. Process

• Source tags are present and honest in every file. [memory] items carry confidence. The arbiter retags the ring-laser 348.6 Hz figure as [web] and says it did not re-fetch it, which is honest.
• Items 18 and 19 above show that two [memory] facts (the free-air gradient and the GPS period) can be retagged [read] from documents in the folder.
• Web checks: the physicist used 4 of 6, the contrarian 4 of 6, and the arbiter about 5 (Qantas, Thirring, Trefethen, Field, WCP5561). The mathematician's [memory] flight time was corrected by the arbiter on the web. The outcome of Trefethen 1965 is honestly marked "secondary".
• Verdict scale: applied consistently. The arbiter's split verdicts each use scale names. "Observation Established / proof of flatness Never properly tested" is clearly explained.
• Primer:
• It follows STEP 4 sections 1-9 in order. Symbols are defined in 3.0. The worked example uses real numbers, and the answers are at the end.
• Reader profile: the acronyms NGA, DMA and BAAS are used undefined (primer lines 3, 65, 262, 265). In full they are the National Geospatial-Intelligence Agency, the Defense Mapping Agency and the British Association for the Advancement of Science.
• Its own "Corrections to earlier documents" (the hidden-height formula; the k-corrected bulge) are right; I confirmed them with example.py Part 4.
• example.py runs and matches the primer.
• Corrections not yet applied. The physicist's file still contains the "10^11 ... 1.2e34 J" energy ratio (line 121) and the wrong heat-budget remark (line 110). The contrarian's file still has "263, not 269" (line 124). The arbiter lists these in §4 (items 1-3), but the source files are unchanged. The orchestrator should apply them before Step 5 entries quote the reviewer files.
• Owner privacy: no owner name or opinion appears in any review file, and the README is private. No defect.

7. Corrections to earlier documents

1. primer.md lines 3 and 262: "R. D. Oldham" should read H. Yule Oldham, M.A., geographer at Cambridge [read: Oldham _text title page; deep read row "Author"; facts.json author "H. Yule Oldham"]. arbiter.md §5 row "Oldham 1900-01": delete "Oldham's papers (Geological Survey of India context)"; the right archive is the Royal Geographical Society or Cambridge geography and the BAAS Section E records.
2. physicist.md claim 1 (c): "Naylor and Carpenter 1870s (cited, not held)" is wrong for Naylor. His August 1894 levelled repeat is in the folder (Earth Review pp.216-217).
3. arbiter.md claim 5 caveat and §7: "The Reddit/Steemit list names about 17 documents; five were read in depth". Should read: the Reddit list names 32 items and the Steemit list 15 (largely overlapping); the folder holds 22 of the reports, 5 deep-read and 17 checked by phrase search and digest. Ten listed items (two CIA, NOAA, IRIG 106-17, NAVAIR and seven ARL reports) are not in the folder.
4. arbiter.md claim 2, difference 4, "Knodel/Campanella ring-gyro press report": the gyroscope was Knodel's. Campanella's was a separate four-mile light-through-holes test that showed curvature (Earth.com). Also in facts.json (Kottke entry): the nature of Campanella's test is described in the Earth.com article in the folder.
5. deep/1893-1897... header: "the Bedford-related items were assigned separately". No such deep read exists; R1 fills the gap.
6. facts.json, Airy 1856 entry: how_read "first of 5 chunks read" is accurate, but the summary should state the result: gravity is greater below by 1/19,286, +2.24 s/day at 1,256 ft [read: _text p.35, p.37].
7. Still pending from the arbiter's §4: physicist line 121 (energy ratio about 5 × 10⁷, not 10¹¹), physicist line 110 (flux is not an objection), contrarian line 124 (269, not 263).

8. Remediation

Deep reads (one round), most important first:

R7 (not a manifest document, so no deep-read slug). The arbiter's QE revision should read and weigh the unread saved files in research/files/: - the Westlake letters of 2 and 9 April 1870; - Carpenter's Field letters of 2 and 9 April 1870; - Hampden's Field letter of 2 April and his Weston letter, which reports "a series of experiments" in the week after; - Chambers's Journal (1870); - The Zetetic (July 1872). Each is a few pages.

What each later file must change

• facts.json:
• Earth Review: add the Naylor 1894 test, the reproduced sketches, Roy's pickets, the heliograph item and Naylor's p.175 concession to key_claims and the summary.
• Koresh: add the optical tests.
• Rowbotham 1865 and 1881: add the vertical-gun, star-tube and sun-altitude experiments.
• Airy 1856: add the result; update how_read after R4.
• IS-GPS: add the clock offset and A_REF.
• Earth.com and Kottke: describe Campanella's test.
• Wikipedia: add Stretton (1905) and Morrow (1896).
• Update how_read for every deep read done.
• claims.md:
• Claim 1 counter-records: add Campanella (2019) and Stretton (1905; secondary).
• Claim 1 flat-side records: add Naylor (1894) and the Koreshan optical tests (1896-97).
• Claim 2: add Rowbotham's vertical-gun and star-tube tests as the inventor's own evidence.
• Coverage note: replace "about 15 pages" for the Review. Correct "Knodel gyroscope drift" as in Corrections item 4.
• arbiter.md (a "QE revision" section):
• Claim 1: weigh Naylor 1894, the Koreshan optical tests, Campanella, Stretton and the heliograph under the existing split verdicts. Add Rowbotham's procedural objection to difference 3 (Retention 1). Name Roy's pickets as historical evidence of variable k. List Middleton's plumb-line design under "settle".
• Claim 2: add Rowbotham's gun experiments with the Coriolis calculation. Verdict: "Coriolis never demonstrated" stands; the inventor's gun test cannot discriminate (about 0.3 m at 16 s).
• Claim 3: quote Naylor's 1895 concession; give the moonlight-temperature claims a verdict line.
• Claim 4: replace the [memory] gradient with Airy 1856 [read]; add the IS-GPS relativistic offset as a second test (mass gravity 4.46 × 10⁻¹⁰ against UA 2.2 × 10⁻⁹).
• §5: correct the Oldham row; add Goudey and Winship as "held as leads, not examined". Correct the claim 5 count.
• Verdicts are expected to stand. If R1 shows Naylor recorded numerical readings inconsistent with a globe at plausible k, claim 1 strong form needs a stated caveat, not a new verdict, unless the readings survive the k analysis.
• primer.md:
• Fix the Oldham name (twice); define NGA, DMA and BAAS.
• In "Kept in the record", add Rowbotham's 1870 objection, Naylor 1894 as the flat side's repeat, Campanella's test and the Koreshan optical sightings (one sentence each, tagged).
• Section 3.5: retag the gradient as [read: Airy 1856] and add the GPS offset as a one-line check.
• Optional for section 8: a check-yourself question on the heliograph (two horizon distances).
• physicist.md, contrarian.md: apply the pending corrections (Process; Corrections item 7) and the Naylor correction (Corrections item 2).
• QE recheck (Sonnet): append "Recheck" after the round.

Recheck

Checked by grep and reading the current files (arbiter.md, primer.md, claims.md, facts.json, deep/, physicist.md, contrarian.md).

The three remaining items are stale errors in reviewer files. They are low risk because arbiter.md section 4 and 9 state the corrections, but they were required before Step 5 entries quote those files.

Final: OPEN (2 issues)

Orchestrator fix (2026-10-10)

The three remaining stale items are corrected in the reviewer files, following the arbiter's corrections: physicist.md claim 1(c) Naylor 1894 now held; physicist.md heat-budget remark replaced by the arbiter's calculation (about 1.8 kW/m^2, no objection); physicist.md energy ratio now about 5 x 10^7 against 7.2 x 10^37 J; contrarian.md Michelson-Gale count now 269. The Flat-Earth entry already avoided these figures.

Final: PASS